TY - GEN
T1 - This Item Might Reinforce Your Opinion
T2 - 32nd ACM Conference on Hypertext and Social Media, HT 2021
AU - Rieger, Alisa
AU - Draws, Tim
AU - Theune, Mariët
AU - Tintarev, Nava
PY - 2021
Y1 - 2021
N2 - During online information search, users tend to select search results that confirm previous beliefs and ignore competing possibilities. This systematic pattern in human behavior is known as confirmation bias. In this paper, we study the effect of obfuscation (i.e., hiding the result unless the user clicks on it) with warning labels and the effect of task on interaction with attitude-confirming search results. We conducted a preregistered, between-subjects crowdsourced user study (N=328) comparing six groups: Three levels of obfuscation (targeted, random, none) and two levels of task (joint, two separate) for four debated topics. We found that both types of obfuscation influence user interactions, and in particular that targeted obfuscation helps decrease interaction with attitude-confirming search results. Future work is needed to understand how much of the observed effect is due to the strong influence of obfuscation, versus the warning label or the task design. We discuss design guidelines concerning system goals such as decreasing consumption of attitude-confirming search results, versus nudging users toward a more analytical mode of information processing. We also discuss implications for future work, such as the effects of interventions for confirmation bias mitigation over repeated exposure. We conclude with a strong word of caution: measures such as obfuscations should only be used for the benefit of the user, e.g., when they explicitly consent to mitigating their own biases.
AB - During online information search, users tend to select search results that confirm previous beliefs and ignore competing possibilities. This systematic pattern in human behavior is known as confirmation bias. In this paper, we study the effect of obfuscation (i.e., hiding the result unless the user clicks on it) with warning labels and the effect of task on interaction with attitude-confirming search results. We conducted a preregistered, between-subjects crowdsourced user study (N=328) comparing six groups: Three levels of obfuscation (targeted, random, none) and two levels of task (joint, two separate) for four debated topics. We found that both types of obfuscation influence user interactions, and in particular that targeted obfuscation helps decrease interaction with attitude-confirming search results. Future work is needed to understand how much of the observed effect is due to the strong influence of obfuscation, versus the warning label or the task design. We discuss design guidelines concerning system goals such as decreasing consumption of attitude-confirming search results, versus nudging users toward a more analytical mode of information processing. We also discuss implications for future work, such as the effects of interventions for confirmation bias mitigation over repeated exposure. We conclude with a strong word of caution: measures such as obfuscations should only be used for the benefit of the user, e.g., when they explicitly consent to mitigating their own biases.
KW - cognitive bias mitigation
KW - confirmation bias
KW - nudging
KW - obfuscation
KW - warning labels
KW - web search
UR - http://www.scopus.com/inward/record.url?scp=85114764718&partnerID=8YFLogxK
U2 - 10.1145/3465336.3475101
DO - 10.1145/3465336.3475101
M3 - Conference contribution
AN - SCOPUS:85114764718
T3 - HT 2021 - Proceedings of the 32nd ACM Conference on Hypertext and Social Media
SP - 189
EP - 199
BT - HT 2021 - Proceedings of the 32nd ACM Conference on Hypertext and Social Media
PB - Association for Computing Machinery (ACM)
Y2 - 30 August 2021 through 2 September 2021
ER -